DirectoryServices crashing since 10.6.3

Question

Level 1

0 points

DirectoryServices crashing since 10.6.3

Hi,
I just upgraded to 10.6.3 on an XServe, and it seemed to work ok. But after a while (maybe an hour) nobody can log on to the server any more. Not ServerAdmin, not ScreenSharing... I have twice now just powered it off, since I didn't see anything else I could do. Anyway, after last restart I just stayed logged in to see if anything happening. The log entries that I assume are the relevant ones:

Apr 18 19:03:19 mail com.apple.DirectoryServices[29]: Assertion failed: (name1), function rbt compare_namekey, file /SourceCache/DirectoryServiceDaemon/DirectoryService-621.3/Server/Mbrd_HashTabl e.c, line 88.
Apr 18 19:03:19 mail com.apple.launchd[1] (com.apple.DirectoryServices[29]): Job appears to have crashed: Abort trap
Apr 18 19:03:19 mail DirectoryService[3610]: Improper shutdown detected
Apr 18 19:03:19 mail servermgrd[62]: servermgr_accounts: noteDirectoryNodeAdded (reopening nodes)
Apr 18 19:03:19 mail com.apple.ReportCrash.Root[3609]: 2010-04-18 19:03:19.696 ReportCrash[3609:2803] Saved crash report for DirectoryService[29] version ??? (???) to /Library/Logs/DiagnosticReports/DirectoryService 2010-04-18-190319localhost.crash

Apr 18 19:03:21 mail servermgrd[62]: servermgr_accounts: got error 5300 trying to auth to local LDAP node
Apr 18 19:03:21 mail servermgrd[62]: servermgr_accounts: noteDirectorySearchPolicyChanged (reopening nodes)
Apr 18 19:03:21 mail servermgrd[62]: servermgr_accounts: got error 5300 trying to auth to local LDAP node
Apr 18 19:03:31 mail ManagedClient[3613]: MCX.MCXD_StartMCX Unable to find matching user name for uid 1000.
Apr 18 19:03:31 mail com.apple.loginwindow[69]: 2010-04-18 19:03:31.487 ManagedClient[3613:903] MCX.MCXD_StartMCX Unable to find matching user name for uid 1000.
Apr 18 19:03:31 mail ManagedClient[3613]: ManagedClient- runloop could not be started; status = -23
Apr 18 19:03:31 mail com.apple.loginwindow[69]: 2010-04-18 19:03:31.489 ManagedClient[3613:903] ManagedClient- runloop could not be started; status = -23

Apr 18 19:38:18 mail servermgrd[62]: servermgr_accounts: got error 5300 trying to auth to local LDAP node
Apr 18 19:38:50 mail com.apple.launchd.peruser.1000[280] (com.apple.PreferenceSyncAgent[4388]): The account "dlorenz" has been deleted out from under us!
Apr 18 19:38:50 mail com.apple.launchd.peruser.1000[280] (com.apple.PreferenceSyncAgent[4388]): In a future build of the OS, this error will be fatal.
Apr 18 19:38:52 mail com.apple.launchd.peruser.1000[280] (com.apple.SubmitDiagInfo[4389]): The account "dlorenz" has been deleted out from under us!
Apr 18 19:38:52 mail com.apple.launchd.peruser.1000[280] (com.apple.SubmitDiagInfo[4389]): In a future build of the OS, this error will be fatal.
Apr 18 19:38:52 mail /System/Library/CoreServices/coreservicesd[104]: scserverServerCheckin: client uid validation failure; getpwuid(1000) == NULL
Apr 18 19:38:58 mail com.apple.launchd.peruser.1000[280] ([0x0-0x12012].com.apple.iCal.helper[4393]): The account "dlorenz" has been deleted out from under us!
Apr 18 19:38:58 mail com.apple.launchd.peruser.1000[280] ([0x0-0x12012].com.apple.iCal.helper[4393]): In a future build of the OS, this error will be fatal.
Apr 18 19:38:58 mail /System/Library/CoreServices/coreservicesd[104]: scserverServerCheckin: client uid validation failure; getpwuid(1000) == NULL

The last bunch about this being fatal reappears every hour from then on...

Any ideas where/what to check?

Thanks,
Dix

Posted on Apr 18, 2010 2:20 PM

Reply

Answer 1

Apr 19, 2010 12:33 AM in response to Dix Lorenz

Hi Dix!

First of all, I don't have a solution but can only contribute another similar situation on our customer's server which does not lead to denial of any user and service, but only to denial of AFP and SMB logins. But this server is "only" a fileserver, so it's nearly every service affected.

The scenario, seen from the aggregated server's log is as follows:

18.04.10 21:50:58 com.apple.backupd[95627] Starting standard backup
18.04.10 21:51:07 com.apple.backupd[95627] Backing up to: /Volumes/TimeMachine/Backups.backupdb
18.04.10 21:51:07 servermgrd[62271] servermgr_backup: TimeMachinePreBackupHook called.
18.04.10 21:51:13 hdiejectd[95679] running
18.04.10 21:51:25 hdiejectd[95679] quitCheck: calling exit(0)
18.04.10 21:52:03 hdiejectd[95837] running
18.04.10 21:52:15 hdiejectd[95837] quitCheck: calling exit(0)
18.04.10 21:52:22 servermgrd[62271] servermgr_backup: TimeMachinePreBackupHook done.
18.04.10 21:52:23 com.apple.backupd[95627] No pre-backup thinning needed: 1.48 GB requested (including padding), 361.26 GB available
*18.04.10 21:52:24 com.apple.DirectoryServices[11425] Assertion failed: (name1), function rbt compare_namekey, file /SourceCache/DirectoryServiceDaemon/DirectoryService-621.3/Server/Mbrd_HashTabl e.c, line 88.*
*18.04.10 21:52:26 com.apple.launchd[1] (com.apple.DirectoryServices[11425]) Job appears to have crashed: Abort trap*
*18.04.10 21:52:26 DirectoryService[95994] Improper shutdown detected*
*18.04.10 21:52:26 com.apple.ReportCrash.Root[95992] 2010-04-18 21:52:26.939 ReportCrash[95992:2a03] Saved crash report for DirectoryService[11425] version ??? (???) to /Library/Logs/DiagnosticReports/DirectoryService 2010-04-18-215226localhost.crash*
18.04.10 21:52:27 com.apple.backupd[95627] Error: (-8084) SrcErr:YES Copying /private/var/db/DirectoryService/flatfile.db-journal to (null)
18.04.10 21:52:37 com.apple.backupd[95627] Copied 445 files (185.4 MB) from volume Server HD.
18.04.10 21:52:37 com.apple.backupd[95627] Copied 448 files (185.4 MB) from volume BOGK Data 1.
18.04.10 21:52:38 com.apple.backupd[95627] No pre-backup thinning needed: 1.24 GB requested (including padding), 361.08 GB available
18.04.10 21:52:40 com.apple.backupd[95627] Copied 133 files (3.7 MB) from volume Server HD.
18.04.10 21:52:40 com.apple.backupd[95627] Copied 136 files (3.7 MB) from volume BOGK Data 1.
18.04.10 21:52:41 servermgrd[62271] servermgr_backup: TimeMachinePostBackupHook called.
18.04.10 21:52:41 servermgrd[62271] servermgr_backup: TimeMachinePostBackupHook done.
18.04.10 21:52:41 com.apple.backupd[95627] Starting post-backup thinning
18.04.10 21:52:43 com.apple.backupd[95627] Deleted backup /Volumes/TimeMachine/Backups.backupdb/BOGK Server 2/2010-04-17-205235: 361.26 GB now available
18.04.10 21:52:43 com.apple.backupd[95627] Post-back up thinning complete: 1 expired backups removed
18.04.10 21:52:43 com.apple.backupd[95627] Backup completed successfully.

Since the problem also reappears here (about once a night -- they always come at night ;-)), I've compared the log entries to the occurrencies before and it's always the same situation: TimeMachine is just preparing its scheduled backup ("TimeMachinePreBackupHook done") and hdiejectd was just called twice.

So maybe you could check if you see the same in your server's Console when selecting "All messages". Another thing: How did you install the 10.6.3 server system? We have several 10.6 Server installations running and the problem only exists on this single machine. The other installations were "grown" from 10.6, then 10.6.1 over 10.6.2 to 10.6.3, but this one was installed from scratch; so we installed 10.6 and applied the 10.6.3 combo update (first version, not 1.1). And only here we have the problem.

When performing a Software Update check, The "Mac OS X 10.6.3 Supplemental Update" is offered, but at the moment I can't apply it since the server is productive. I will run it this evening, maybe it helps!

Ah, by the way -- With our "problem" server we could fix the issue rather simple: I just have to stop and start over the AFP service with Server Admin to get things working again. All users can login then until next time the DirectoryServices have crashed. But I fear this won't help you as you are not able to authenticate at all to your server ...

Cheers,
Ron.

Reply

Answer 2

Apr 20, 2010 4:28 PM in response to Dix Lorenz

Dix Lorenz:
Are you doing time machine backups of your server?
If you're still seeing the DirectoryService crash could you post the crashing thread (/Library/Logs/CrashReporter)

Reply

Answer 3

davidh

Level 4

1,890 points

Apr 20, 2010 7:54 PM in response to David Kurtz2

The answer is right there in the log entry:

18.04.10 21:52:27 com.apple.backupd95627 Error: (-8084) SrcErr:YES Copying /private/var/db/DirectoryService/flatfile.db-journal to (null)

You cannot, must NOT, ever, backup a database externally to itself while that database is in use.
Barring other OS'es and filesystems that potentially allow for other means for this (and still, it's not advisable)

At best, you might possibly get a "backup" that is marked as inconsistent (internally within the database) and would require a rebuild anyway in order to use - if possible and available.

Meanwhile, by trying to backup a DB externally, you seriously risk interrupting normal (internal) operations to/for that database that is in operation.

If you must use Time Machine to backup Mac OS X Server, you do need to make certain to exclude ALL directories with any form of database in them. This includes those that are a part of OpenLDAP, local node db's in /var/db, and numerous sqlite DBs used by other services such as mail,
within /Library/CalendarServer, and so on.

You'd do well to exclude all of /var because of how actively that's constantly in use by the OS anyway.

Reply

Answer 4

Dix Lorenz Author

Level 1

0 points

Apr 21, 2010 12:06 AM in response to davidh

Thanks for all the replies...

Indeed I am backing up the Server via TimeMachine. I have turned it off 2 days ago, since then the Server has been running without any problems.

davidh wrote:
If you must use Time Machine to backup Mac OS X Server, you do need to make certain to exclude ALL directories with any form of database in them. This includes those that are a part of OpenLDAP, local node db's in /var/db, and numerous sqlite DBs used by other services such as mail,
within /Library/CalendarServer, and so on.

You'd do well to exclude all of /var because of how actively that's constantly in use by the OS anyway.

I understand the reasoning behind it and to be honest, I always wondered how a backup of a live db could possibly work. 2 questions remain in my mind though:

How does one backup a server then?
Why did it work in 10.6.2 but not 10.6.3?

Thanks,
Dix

Reply

Answer 5

Dix Lorenz Author

Level 1

0 points

Apr 21, 2010 12:07 AM in response to David Kurtz2

Here's the crashing thread in DirectoryService:

Thread 7 Crashed: Dispatch queue: User Name
0 libSystem.B.dylib 0x00007fff85a7f01e _semwait_signalnocancel + 10
1 libSystem.B.dylib 0x00007fff85a7ef20 nanosleep$NOCANCEL + 129
2 libSystem.B.dylib 0x00007fff85adb902 usleep$NOCANCEL + 57
3 libSystem.B.dylib 0x00007fff85afaeb8 abort + 93
4 libSystem.B.dylib 0x00007fff85ae7ef0 _pthreadmarkcancel + 0
5 DirectoryService 0x00000001000a16f1 rbt compare_namekey + 68
6 DirectoryService 0x00000001000d3b08 rb tree_findnode + 38
7 DirectoryService 0x00000001000a1604 _HashTable_GetAndRetain_block_invoke1 + 30
8 libSystem.B.dylib 0x00007fff85a4189b dispatch barrier_syncf + 79
9 DirectoryService 0x00000001000a12d6 HashTable_GetAndRetain + 126
10 DirectoryService 0x00000001000d2de9 MbrdCache_GetAndRetain + 476
11 DirectoryService 0x00000001000baadc _Mbrd_GetItemWithIdentifierAndRetain(MbrdCache*, int, void const*, int) + 440
12 DirectoryService 0x00000001000bb016 _Mbrd_GetItemWithIdentifierAndRetain_block_invoke4 + 72
13 libSystem.B.dylib 0x00007fff85a4189b dispatch barrier_syncf + 79
14 DirectoryService 0x00000001000b50cd Mbrd GetItemWithIdentifierAndRetain(MbrdCache*, int, void const*, int) + 201
15 DirectoryService 0x00000001000b582f Mbrd_ProcessMapIdentifier + 848
16 DirectoryService 0x00000001000299b5 memberdDSmig doMapName + 268
17 DirectoryService 0x00000001000a0e35 _XMapName + 233
18 DirectoryService 0x00000001000a08ca DSmemberdMIG_server + 106
19 DirectoryService 0x0000000100028a85 dsmig demux_internaldispatch(mach_msg_headert*, mach msg_headert*) + 45
20 libSystem.B.dylib 0x00007fff85a890dc mach msgserver + 357
21 libSystem.B.dylib 0x00007fff85a48610 dispatch_call_block_andrelease + 15
22 libSystem.B.dylib 0x00007fff85a26b90 dispatch_workerthread2 + 206
23 libSystem.B.dylib 0x00007fff85a264e8 pthreadwqthread + 353
24 libSystem.B.dylib 0x00007fff85a26385 start_wqthread + 13

Thread 7 crashed with X86 Thread State (64-bit):
rax: 0x000000000000003c rbx: 0x0000000100709740 rcx: 0x00000001007096f8 rdx: 0x0000000000000001
rdi: 0x0000000000000c03 rsi: 0x0000000000000000 rbp: 0x0000000100709730 rsp: 0x00000001007096f8
r8: 0x0000000000000000 r9: 0x0000000000989680 r10: 0x0000000000000001 r11: 0xffffff80002e1670
r12: 0x0000000000000000 r13: 0x00000001000ed0b8 r14: 0x00000001000ed16b r15: 0x00007fff709d2670
rip: 0x00007fff85a7f01e rfl: 0x0000000000000247 cr2: 0x00000001001fe000

Reply

Answer 6

davidh

Level 4

1,890 points

Apr 21, 2010 2:26 AM in response to Dix Lorenz

Consider it a necessary wake-up call (to you) 🙂

Effectively, it would appear that your backup attempt is causing a wait for DirectoryService - and within/as a part of that, Mbrd (memberd), see
http://www.opensource.apple.com/source/DirectoryService/DirectoryService-621.1/S erver/Mbrd_Cache.h

and "DSmemberdMIG_server" in
http://www.opensource.apple.com/source/DirectoryService/DirectoryService-621.3/S erver/ServerControl.cpp

At a glance, I'd say that Time Machine is throwing a wrench into the gears of DirectoryService/memberd and - as should be expected - that's causing a problem.

See also semwait ,
http://publib.boulder.ibm.com/infocenter/zvm/v5r3/index.jsp?topic=/com.ibm.zvm.v 53.dmsd0/semwt.htm

http://en.wikipedia.org/wiki/Semaphore_(programming)

Reply

Answer 7

mawt

Level 1

0 points

Apr 22, 2010 11:03 PM in response to Dix Lorenz

Hi,
I have a similar problem, running 10.6.3 on an XServe, users are kicked out (can't access AFP, can't login) every so often (twice a day). Looking at the console showed it was happening around Time Machine was doing a backup, but it wasn't happening every time there was a backup. And I am only backing up the users' AFP files, along with user directories (ie no databases that I can see).

So, I am turning off Time Machine until I'm guaranteed a solution. But I am considering the following culprits
- hidden files on our AFP Server (eg spotlight files)
- something in the users' directories
- the inability of Time Machine to send an email

Could one of these be the culprit?

Console Messages:
23/04/10 12:24:57 PM com.apple.backupd[7425] Starting standard backup
23/04/10 12:24:57 PM com.apple.backupd[7425] Backing up to: /Volumes/Archive2/Backups.backupdb
23/04/10 12:24:57 PM servermgrd[74] servermgr_backup: TimeMachinePreBackupHook called.
23/04/10 12:25:14 PM hdiejectd[7480] running
23/04/10 12:25:27 PM hdiejectd[7480] quitCheck: calling exit(0)
23/04/10 12:25:52 PM com.apple.DirectoryServices[29] Assertion failed: (name1), function rbt compare_namekey, file /SourceCache/DirectoryServiceDaemon/DirectoryService-621.3/Server/Mbrd_HashTabl e.c, line 88.
23/04/10 12:25:53 PM com.apple.launchd[1] (com.apple.DirectoryServices[29]) Job appears to have crashed: Abort trap
23/04/10 12:25:53 PM DirectoryService[7677] Improper shutdown detected
23/04/10 12:25:53 PM com.apple.ReportCrash.Root[7675] 2010-04-23 12:25:53.114 ReportCrash[7675:2903] Saved crash report for DirectoryService[29] version ??? (???) to /Library/Logs/DiagnosticReports/DirectoryService 2010-04-23-122553localhost.crash
23/04/10 12:25:57 PM hdiejectd[7689] running
23/04/10 12:26:11 PM hdiejectd[7689] quitCheck: calling exit(0)
23/04/10 12:26:26 PM servermgrd[74] servermgr_backup: TimeMachinePreBackupHook done.
23/04/10 12:26:30 PM com.apple.backupd[7425] No pre-backup thinning needed: 1.05 GB requested (including padding), 3.47 TB available
23/04/10 12:26:35 PM mdworker32[7856] kCGErrorFailure: Set a breakpoint @ CGErrorBreakpoint() to catch errors as they are logged.
23/04/10 12:27:07 PM com.apple.backupd[7425] Copied 2052 files (11.0 MB) from volume RAID.
23/04/10 12:27:07 PM com.apple.backupd[7425] No pre-backup thinning needed: 1.04 GB requested (including padding), 3.47 TB available
23/04/10 12:27:09 PM com.apple.backupd[7425] Copied 232 files (0 bytes) from volume RAID.
23/04/10 12:27:10 PM servermgrd[74] servermgr_backup: TimeMachinePostBackupHook called.
23/04/10 12:27:10 PM servermgrd[74] servermgr_backup: TimeMachinePostBackupHook done.
23/04/10 12:27:11 PM com.apple.backupd[7425] Starting post-backup thinning
23/04/10 12:27:15 PM com.apple.backupd[7425] Deleted backup /Volumes/Archive2/Backups.backupdb/2010-04-22-120113: 3.47 TB now available
23/04/10 12:27:15 PM com.apple.backupd[7425] Post-back up thinning complete: 1 expired backups removed
23/04/10 12:27:15 PM com.apple.backupd[7425] Backup completed successfully.
23/04/10 12:31:57 PM servermgrd[74] [74] error in getAndLockContext: flock returned -1 (errno=9) fd = 41 file was /private/var/servermgrd//servermgr_ipfilter.lock
23/04/10 12:34:38 PM com.apple.wikid.mailinglists[8196] mail:status = "list-updated"
23/04/10 12:44:38 PM com.apple.wikid.mailinglists[8539] mail:status = "list-updated"

Reply

Answer 8

roossym

Level 1

5 points

Apr 23, 2010 4:48 AM in response to Dix Lorenz

Exactly same problem here since update of SLS to 10.6.3.

I'm really curious if Apple ever test their server stuff.
There was the OD db transaction log bug before 10.6.3 (is it fixed with 10.6.3?) which crashed OD after 4 days, now we have this new bug. For a SERVER OS, this is a shame.

Thread 6 Crashed: Dispatch queue: User Name
0 libSystem.B.dylib 0x00007fff87fff8c9 usleep$NOCANCEL + 0
1 libSystem.B.dylib 0x00007fff8801eeb8 abort + 93
2 libSystem.B.dylib 0x00007fff8800bef0 _pthreadmarkcancel + 0
3 DirectoryService 0x00000001000a16f1 rbt compare_namekey + 68
4 DirectoryService 0x00000001000d3b08 rb tree_findnode + 38
5 DirectoryService 0x00000001000a1bd9 _HashTableAdd + 838
6 DirectoryService 0x00000001000a205b _HashTable_Add_block_invoke1 + 34
7 libSystem.B.dylib 0x00007fff87f6589b dispatch barrier_syncf + 79
8 DirectoryService 0x00000001000a1396 HashTable_Add + 142
9 DirectoryService 0x00000001000d2722 MbrdCache AddToHashes(MbrdCache*, UserGroup*) + 260
10 DirectoryService 0x00000001000d3322 MbrdCache_AddOrUpdate + 962
11 DirectoryService 0x00000001000b9aa6 Mbrd_FindItemsAndRetain(unsigned int, tDataList*, int, char const*, unsigned int, unsigned int*) + 4549
12 DirectoryService 0x00000001000ba23b Mbrd_FindItemAndRetain(unsigned int, tDataList*, int, char const*, unsigned int) + 47
13 DirectoryService 0x00000001000ba3af _Mbrd_GenerateItemMembership_block_invoke8 + 245
14 libSystem.B.dylib 0x00007fff87f6c610 dispatch_call_block_andrelease + 15
15 libSystem.B.dylib 0x00007fff87f4b1b2 dispatch_queuedrain + 251
16 libSystem.B.dylib 0x00007fff87f4b014 dispatch_queueinvoke + 57
17 libSystem.B.dylib 0x00007fff87f4abbe dispatch_workerthread2 + 252
18 libSystem.B.dylib 0x00007fff87f4a4e8 pthreadwqthread + 353
19 libSystem.B.dylib 0x00007fff87f4a385 start_wqthread + 13

Reply

Answer 9

Apr 25, 2010 2:00 AM in response to roossym

Hi all!

Here is my current situation (I described the situation on our customer's server some posts above): As I’ve mentioned in another thread, I also noticed that the problem came up when applying 10.6.3 combo (first release) onto an out-of-the-box installation of SLS 10.6.0. We have some more customers using 10.6.3 Server with "grown" systems, so they were updated from 10.6.0 to 10.6.1 to 10.6.2 to 10.6.3. They never were affected by this problem, having more users and more load (and also TM backups of the server system running).

So I tried applying the 10.6.3v1.1 combo updater to the affected system and since then (five days ago) everything appears to be fine. No more DS crashes, no more AFP or SMB problems or restarts.

Regarding the TimeMachine as a potential problem: Of course backing up living database files is a problem and as far as I know that should be handled by TM running on a server. Since there definitely were massive problems under 10.5 Server (best example was the Cyrus Mailserver under Leopard), Apple does not backup certain directories now and takes care that any live backup of a system database is secure. Obviously there is a problem with 10.6.3 (maybe only 10.6.3v1), but I would not generally disable TM backups of the server system (of course as far as they are stable ...).

And I can only repeat Dix' questions:

Dix Lorenz wrote:
How does one backup a server then?
Why did it work in 10.6.2 but not 10.6.3?

So since the problem seems to be related to (a) 10.6.3 (b) running OD DirectoryServices and (c) TimeMachine (obviously independent from backing up DirectoryServices' databases, see mawt!), what would be interesting to me is: Do the affected systems have a "grown" system as described above or was the original 10.6.3 combo update applied to a 10.6.0 install?

And mawt: Did you really explicitly exclude any system folder (e.g. /var) or the complete startup volume from your TimeMachine backup?

Cheers,
Ron.

Reply

Answer 10

Apr 25, 2010 11:29 PM in response to davidh

Hi all, me (again) ...

As often, when talked about it a problem suddenly wakes up again: Yesterday I've reported that after the 10.6.3v1.1 update six days ago, DirectoryServices did not hang up again. Well, they did. An hour ago. Business as usual, happening during the start of the backup, stop/start AFP solves the issue.

So now we know, that it's not connected to 10.6.3v1/10.6.3v1.1 but it is a general 10.6.3 problem (which did not appear under earlier 10.6 Servers. I now explicitly excluded /private/var/db from the TimeMachine backup and will report again ...

Cheers,
Ron.

Reply

Answer 11

etblack

Level 1

18 points

Apr 26, 2010 6:34 PM in response to Ronald Altmann

I have run into the same problem on two systems. On one I rebuilt the OD database, exporting users and groups, demoting to standalon, recreating as OD Master and reimporting and resetting passwords. Since then the first one hasn't crashed so I thought the procedure would work fine on the second (these are both systems with fewer than 10 users so it isn't as big as a put out). However, the second crashed again today after a week of behaving itself.

Trying to exclue /var/db

Reply

Answer 12

davidh

Level 4

1,890 points

Apr 26, 2010 7:16 PM in response to etblack

There's no indication that Time Machine is at all appropriate for backing up OD.

In order to do so it would have to stop slapd and then start it again after TM was done backing up.

So yes, exclude /var/db
you may have to use the Finder's Go menu, and select "Go to Folder..." and type in /var
and select within there: db

Still, do keep in mind that anything else you forget to exclude that uses a database (plenty of services do), is asking for a failed backup. IE - db's in your backup will be in an inconsistent state and would need to be rebuilt before use - not at all a desirable situation, something to be avoided with any proper practice(s)

Reply

Answer 13

offdwall

Level 1

4 points

Apr 27, 2010 2:42 PM in response to etblack

I have had this issue of OD crashing every day or two. After reading a few threads here, I narrowed it down to occurring right after a Time Machine backup of the server. Not after every TM BU but every two days or so. Sometimes a reboot will fix it but most times the server HD must be restored from the last TM BU. I don't buy into the fact that TM can't back up OD because the restore ALWAYS produces a fully functional OSX server. The issue seems to be that the act of finishing up the TM backup crashes OD after a successful BU has been performed. I've tried excluding /var and that has no effect. I even tried excluding the entire system drive. Just running a TM BU of the separate SATA data drive will crash OD after a day or so.

I've just disabled Time Machine completely on the server and it has been running for two days with no OD crashes (a new record since 10.6.3). If it goes another day, I'll be convinced that there is a serious bug with OSX Server 10.6.3 and Time Machine.

Reply

Answer 14

Apr 27, 2010 3:02 PM in response to offdwall

Hi again!

Thank you, offdwall, for your post! I've also tried to solve the 10.6.3 problem by excluding /var/db from the TimeMachine Backup as I've posted earlier. Just as I saw your reply I took another look on the affected server and — coincidence or not: DirectoryServices have crashed again (assertion failure) after more than one day without problems. Wow!

So it really appears that excluding the DS database from the backup indeed does not solve the problem. I now totally disabled TM and restarted the server (as in our case we did not need to recover the whole system to get the DirectoryService to run ...). I will have a look at it and be reporting ... Also I will try to open a support ticket at Apple's Server support. This is ridiculous!

Cheers,
Ron.

Reply

Answer 15

Apr 27, 2010 3:42 PM in response to Ronald Altmann

I'm having the exact same problem where Time Machine appears to be causing a Open Directory crash. I've also noticed that at almost the exact time Time Machine begins, my slapconfig.log is reporting that slapconfig -backupdb is being called, and under it, it's reporting that it's stopped and restarted the LDAP server. Not sure if this is related or not, but it's possible, as I've never configured slapconfig to run automatically.

Reply